OpenAI
gpt-4o-mini-2024-07-18
Made by OpenAI. Released 2024-07-18. Accessibility not recorded. Cheapest standard rate we hold: $1.20 per million output tokens, no context split.
What it costs
| Tier | Kind | In | Out | Cached in | From |
|---|---|---|---|---|---|
| batch | text | $0.15 | $0.60 | $0.07 | seller |
| standard | text | $0.30 | $1.20 | $0.15 | seller |
per million tokens, US dollars. * marks a rate from a single registry or two that disagree — the best available number rather than a confirmed one. Every other row is confirmed against a seller or a licensed index.
What it scores, by house
| Benchmark | House | Score | Evaluated | Scaffold |
|---|---|---|---|---|
| chess puzzles | Epoch AI | 0.0±0.00 | 2026-07-15 | — |
| frontiermath tiers 1 3 v2 | Epoch AI | 0.0±0.00 | 2026-08-27 | — |
| gpqa diamond | Epoch AI | 0.4±0.02 | 2025-01-27 | — |
| math level 5near its ceiling | Epoch AI | 0.5±0.01 | 2025-01-27 | — |
| mystery game puzzles | Epoch AI | 0.1±0.03 | 2026-08-27 | — |
| otis mock aime 2024 2025near its ceiling | Epoch AI | 0.1±0.03 | 2025-07-30 | — |
| simpleqa verified | Epoch AI | 0.1±0.01 | 2026-08-31 | — |
| aider polyglot | collected by Epoch AI | 3.6 | — | — |
| arc agi 2 | collected by Epoch AI | 0.0 | — | — |
| balrog | collected by Epoch AI | 0.2 | — | — |
| bool q | collected by Epoch AI | 0.9 | — | — |
| dtbenchnear its ceiling | collected by Epoch AI | 0.5 | — | — |
| geobench | collected by Epoch AI | 3401 | — | — |
| gsm8k | collected by Epoch AI | 0.9 | — | — |
| lech mazur writing | collected by Epoch AI | 6.7 | — | — |
| live bench | collected by Epoch AI | 41.3 | — | — |
| lmca | collected by Epoch AI | 10.4 | — | — |
| mmlu | collected by Epoch AI | 0.8 | — | — |
| piqa | collected by Epoch AI | 0.9 | — | — |
| simplebench | collected by Epoch AI | 0.1 | — | — |
| video mme | collected by Epoch AI | 0.6 | — | — |
| vpct | collected by Epoch AI | 0.3 | — | — |
| weirdml | collected by Epoch AI | 0.1 | — | — |
These scores are not comparable between houses and are not added up. A score is dated by when it was evaluated, not when the model was released, and a score without its scaffold is not a measurement — which is why both are printed.
Every rate we hold · Why there is no single best model · Everything from OpenAI